Dynamic electronic data generation through human-guided multi-agent collaboration
A multi-agent framework for generating synthetic data includes a human-in-the-loop agent that receives schema definitions, optional seed data, and domain-specific context, and transforms such inputs into a structured schema JSON; and an orchestrator agent that decomposes electronic data generation into subtasks and invokes autonomous agents including a topic modelling agent, an input enrichment agent, a variability injection agent, and a data generation agent. The topic modelling agent employs a large language model with an extended context window to extract hierarchical categories and salient keywords. The input enrichment agent expands categories using retrieval, keyword expansion, and iterative refinement, while the variability injection agent produces structured variation artifacts simulating real-world irregularities. The data generation agent synthesizes multimodal datasets that are persisted by an artifact serializer/loader and validated by a validation tool for coverage, schema conformity, and distributional balance. The synthetic dataset is suitable for training and validating machine learning models.
1 . A system for dynamically generating synthetic data, including a memory and one or more processors, wherein the memory stores a computer program executable by the one or more processors, and when executing the computer program, the one or more processors are configured to perform:
receiving, via a human-in-the-loop interface, input comprising a target data schema for an electronic data generation process;
transmitting the received input to an orchestrator agent configured to decompose the electronic data generation process into discrete tasks and determine an execution strategy based on the discrete tasks; and
invoking, according to the execution strategy, a plurality of autonomous agents to complete the discrete tasks to generate the synthetic data, wherein the plurality of autonomous agents includes:
an input enrichment agent that broadens a number and types of input parameters associated with the received input to generate a first intermediate output;
a variability injection agent that introduces controlled complexity to the received input to generate a second intermediate output; and
a data generation agent that generates the synthetic data conforming to the target data schema by integrating one or more of the first intermediate output and the second intermediate output into the electronic data generation process,
wherein the orchestrator agent provides signed artifact pointers and control payloads to the plurality of autonomous agents instead of raw context data.
2 . The system of claim 1 , wherein the received input further includes seed data, and the plurality of autonomous agents further includes a topic modeling agent configured to perform topic modeling and categorization on the seed data and relevant context for the electronic data generation process, so as to extract hierarchical structures of categories and keywords relevant to the target data schema.
3 . The system of claim 2 , wherein the data generation agent generates the synthetic data conforming to the target data schema by integrating one or more of the first intermediate output, the second intermediate output, and hierarchical structures of categories and keywords relevant to the target data schema.
4 . The system of claim 2 , wherein the topic modeling agent uses a large language model with an extended context window to analyze the seed data.
5 . The system of claim 1 , wherein the plurality of autonomous agents further includes a human-in-the-loop agent that transforms the input, received via the human-in-the-loop interface, into a structured schema in JSON format, the structured schema including separate fields for (i) desired output data schema and (ii) business information and contextual parameters to be supplied to different agents for execution of the discrete tasks, wherein the target data schema specifies a structure and attributes of the synthetic data.
6 . The system of claim 1 , wherein the plurality of autonomous agents further includes a validation tool that applies statistical and structural analyses to ensure that the generated synthetic data meets predefined standards of coverage, variety, and fidelity.
7 . The system of claim 1 , wherein the orchestrator agent operates as a stateful controller that performs capacity planning to determine which agents are required, in what order, and which can run concurrently when the plurality of autonomous agents complete the discrete tasks.
8 . The system of claim 1 , wherein the first intermediate output is an enriched inputs JSON artifact annotated with provenance metadata and relevance scores.
9 . The system of claim 1 , wherein the second intermediate output is a structured variations JSON artifact specifying variation types, frequency parameters, and distribution weights.
10 . The system of claim 1 , wherein the plurality of autonomous agents further include an artifact serializer/loader that externalizes the first intermediate output and the second intermediate output to avoid context-window limitations of large language models.
11 . The system of claim 1 , wherein the data generation agent invokes one or more of a text generation model, an image generation model, or a video generation model to create one or more of a text dataset, an image dataset, or a video dataset in the electronic data generation process.
12 . A system for dynamically generating synthetic data, including a memory and one or more processors, wherein the memory stores a computer program executable by the one or more processors, and when executing the computer program, the one or more processors are configured to perform:
receiving, via a human-in-the-loop interface, input comprising a target data schema for an electronic data generation process;
transmitting the received input to an orchestrator agent configured to decompose the electronic data generation process into discrete tasks and determine an execution strategy based on the discrete tasks; and
invoking, according to the execution strategy, a plurality of autonomous agents to complete the discrete tasks to generate the synthetic data, wherein the plurality of autonomous agents includes:
an input enrichment agent that broadens a number and types of input parameters associated with the received input to generate a first intermediate output;
a variability injection agent that introduces controlled complexity to the received input to generate a second intermediate output; and
a data generation agent that generates the synthetic data conforming to the target data schema by integrating one or more of the first intermediate output and the second intermediate output into the electronic data generation process,
wherein the plurality of autonomous agents further includes a human-in-the-loop agent that transforms the input, received via the human-in-the-loop interface, into a structured schema in JSON format, the structured schema including separate fields for (i) desired output data schema and (ii) business information and contextual parameters to be supplied to different agents for execution of the discrete tasks, wherein the target data schema specifies a structure and attributes of the synthetic data, and
wherein the human-in-the-loop agent defaults to a domain-agnostic mode when no seed data is included in the received input.
13 . A system for dynamically generating synthetic data, including a memory and one or more processors, wherein the memory stores a computer program executable by the one or more processors, and when executing the computer program, the one or more processors are configured to perform:
receiving, via a human-in-the-loop interface, input comprising a target data schema for an electronic data generation process;
transmitting the received input to an orchestrator agent configured to decompose the electronic data generation process into discrete tasks and determine an execution strategy based on the discrete tasks; and
invoking, according to the execution strategy, a plurality of autonomous agents to complete the discrete tasks to generate the synthetic data, wherein the plurality of autonomous agents includes:
an input enrichment agent that broadens a number and types of input parameters associated with the received input to generate a first intermediate output;
a variability injection agent that introduces controlled complexity to the received input to generate a second intermediate output; and
a data generation agent that generates the synthetic data conforming to the target data schema by integrating one or more of the first intermediate output and the second intermediate output into the electronic data generation process,
wherein the plurality of autonomous agents further includes a validation tool that applies statistical and structural analyses to ensure that the generated synthetic data meets predefined standards of coverage, variety, and fidelity, and
wherein the validation tool further outputs a structured feedback report used by the orchestrator agent to re-engage the input enrichment agent or the variability injection agent for refinement.
14 . A computer-implemented method for dynamically generating synthetic data, the method comprising:
receiving, via a human-in-the-loop interface, input comprising a target data schema for an electronic data generation process;
transmitting the received input to an orchestrator agent configured to decompose the electronic data generation process into discrete tasks and determine an execution strategy based on the discrete tasks; and
invoking, according to the execution strategy, a plurality of autonomous agents to complete the discrete tasks to generate the synthetic data, wherein the plurality of autonomous agents includes:
an input enrichment agent that broadens a number and types of input parameters associated with the received input to generate a first intermediate output;
a variability injection agent that introduces controlled complexity to the received input to generate a second intermediate output; and
a data generation agent that generates the synthetic data conforming to the target data schema by integrating one or more of the first intermediate output and the second intermediate output into the electronic data generation process,
wherein the plurality of autonomous agents further includes an artifact serializer/loader that externalizes the first intermediate output and the second intermediate output to avoid context-window limitations of large language models.
15 . The method of claim 14 , wherein the received input further includes seed data, and the plurality of autonomous agents further includes a topic modeling agent configured to perform topic modeling and categorization on the seed data and relevant context for the electronic data generation process, so as to extract hierarchical structures of categories and keywords relevant to the target data schema.
16 . The method of claim 15 , wherein the data generation agent generates the synthetic data conforming to the target data schema by integrating one or more of the first intermediate output, the second intermediate output, and hierarchical structures of categories and keywords from the first intermediate output and the second intermediate output and relevant to the target data schema.
17 . The method of claim 14 , wherein the plurality of autonomous agents further includes a human-in-the-loop agent that transforms the input, received via the human-in-the-loop interface, into a structured schema in JSON format, the structured schema including separate fields for (i) desired output data schema and (ii) business information and contextual parameters to be supplied to different agents for execution of the discrete tasks, wherein the target data schema specifies a structure and attributes of the synthetic data.
18 . The method of claim 14 , wherein the plurality of autonomous agents further includes a validation tool that applies statistical and structural analyses to ensure that the generated synthetic data meets predefined standards of coverage, variety, and fidelity.
19 . The method of claim 18 , further comprising using the synthetic data to train a machine learning model in combination with a limited set of real data.